基于改进多元线性回归分析的雾霾天绝缘子故障诊断研究

Study on Fault Diagnosis of Insulator in Haze Based on Improved Multivariate Linear Regression Analysis

摘要: 本文主要是在模拟雾霾的情形下对绝缘子故障情况进行分析，通过上述实验得到不同情形下产生闪络现象的电流电压数据。进而针对本实验中的特殊性，结合EIV模型的统计学特性，对传统的多元线性回归模型进行一定的改进，使之能更好、更便利地分析各个因变量之间的相互关联性及模型拟合效果，克服了数据稳定性的限制。利用改进后的回归模型，分别建立故障诊断预测模型与小波分析参数区间预测误差模型，进而根据回归分析得到的绝对平均误差和拟合优化检验，结合这两个模型，来判断故障类别，经过对三种不同情形下故障的检验，这两种模型在实际应用中均能够满足要求。

Abstract: The insulator is analyzed under the condition of simulated haze in this paper, and the current and voltage data of the flashover under different conditions are obtained through the above experiments. According to the particularity of the experiment and the statistical characteristics of the Errors in Variables model, the traditional multiple linear regression model has been improved to better and more conveniently analyze the correlation between the various dependent variables and the fitting effect of the model, overcoming the limitation of data stability at the same time. Using the improved regression model, we established the fault diagnosis prediction model and the wavelet analysis parameter interval prediction model respectively, and the fault categories are judged based on the absolute average error and the fitting optimization test obtained by the two models and the regression analysis. After testing, two models can meet the requirements of diagnostic faults and prediction errors.